How Alexa for Shopping (Rufus, COSMO) is rewriting Amazon PPC.

Amazon retired the standalone Rufus assistant in May 2026 and folded it into Alexa for Shopping — and the shift changes Amazon PPC from ranking for a word to winning a conversation. With an AI assistant now mediating an estimated 15–20% of mobile search queries and interpreting intent instead of matching keywords, keyword strategy still matters — but it is no longer the whole game. Here is what actually changes, and what does not.

What is Alexa for Shopping (formerly Rufus)?

Rufus was Amazon's generative shopping assistant, built on Amazon Bedrock, that let shoppers ask questions and get recommendations in natural language. As of 2026 it had over 250 million users with interactions up more than 210% year over year. In May 2026 Amazon retired the standalone Rufus brand and merged its product knowledge into Alexa for Shopping, combining it with Alexa+'s personalisation. Underneath sits COSMO — Amazon's commonsense knowledge framework that infers why a shopper wants something from co-buy and search-buy behaviour, rather than matching the literal words.

How this changes PPC keyword strategy

The headline shift: COSMO interprets intent, so the relationship between a typed query and the products surfaced is looser and more contextual than exact-match keyword logic. Three practical consequences:

• Long-tail and conversational queries grow. Shoppers ask the assistant full questions ("a quiet humidifier for a baby's room under €50"), so the intent behind a purchase is richer than the head term it used to collapse into.
• Context quality competes with keyword coverage. Visibility now also depends on use-case clarity, complete information, ratings, and category relevance — the signals COSMO reads to decide you are the right answer.
• Relevance beats brute-force bidding. Bidding hard on a term you are not genuinely the best answer for gets more expensive, because the AI layer is less likely to reinforce the match.

What does NOT change

It is easy to over-rotate. Keywords are not dead — they are one layer of several. Sponsored Products still wins high-intent placements. Break-even CPC discipline still decides whether a click is worth it. Inventory still constrains how hard you should push. The fundamentals in CPC benchmarks and profit-aware bidding hold. What changes is that keyword coverage alone is no longer sufficient — it is necessary but not the whole strategy.

How to optimise for the COSMO era

1. Write listings that state the use case plainly. Who is it for, what problem does it solve, in what context. COSMO reads this; so do the off-Amazon LLMs shoppers now research in.
2. Cover the question, not just the keyword. Anticipate the natural-language queries around your product and make sure your content answers them completely.
3. Keep semantic relevance tight. Bid where you are genuinely a strong answer. Mirox's semantic engine scores relevance per term so spend concentrates where intent and product actually match — see how the Semantic agent works.
4. Localise per marketplace. COSMO reads intent in the shopper's language; machine-translated copy reads as machine-translated. This compounds the cross-border penalty in EU-native is not a translation layer.

The bottom line

The assistant era rewards the same thing good PPC always rewarded — being genuinely the best answer to a real shopper's need — and punishes the workaround that used to paper over a weak match: bidding harder on a keyword you did not deserve. Relevance is now a budget line, not just a content nicety.

A profit model that scores semantic relevance on every bid, in every marketplace, and shows the score in a trace, is built for exactly this shift. See a trace, or watch it decide on your account in Simulation Mode.